[Report] Google Cloud Next ’20 – Week 8 Cloud AI

Google Cloud Next

“Google Cloud Next” is a conference of “Google Cloud”. It was conducted online in 2020.
In this survey, we will participate in the session about Week 8: Cloud AI of Google Cloud Next ’20 and write the report.

Sponsored Links

Summary of this report

Google Cloud Next is a Google Cloud conference that has been held every year. It was conducted online in 2020.
Google Cloud Next ’20 has been held for 9 weeks, with sessions on different themes every week.

Week 8 had a session on Cloud AI. We attended the session of this conference. Among them, we will report on the sessions conducted as keynotes.

Overview of Solution Keynote – Generating Value with AI

The solution keynote for “Week 8: Cloud AI” is “Generating Value with AI.” You can check the video of this session from the following.

In this session, they show how to use AI to innovate and create value. They also explain how customers can use AI to transform their organizations by using Google Cloud AI. The speaker of this session is Rajen Sheth, Vice President of Google Cloud AI and Industry Solutions at Google.

Details and Opinions of “Solution Keynote – Generating Value with AI”

We describe the contents of this session, including our opinions.

  • Customers can use AI to create value and transform their organizations. The means is to solve the following two problems.
    • Solve common problems with AI
    • Solve unique problems with AI
  • Solution for “Solve common problems with AI”
    • Contact Center AI  (CCAI)
      • https://cloud.google.com/solutions/contact-center
      • This service uses world-class interactive AI.
      • Agent Assist for Chat” announced
        • It supports agents in chat in addition to voice calls
      • “Custom Voice” announced
        • It can create original voice for virtual agents.
      • “Dialogflow CX” announced
        • “Dialogflow” is the core technology of CCAI.
        • It optimizes Dialogflow for large contact centers.
        • It supports complex interactive architecture.
      • [Opinion]
        “Dialoflow CX” allows easy composition of complex dialogue stories. In addition, the interface will be expanded with “Custom Voice” and “Agent Assist for Chat”. We think that call centers can be fully delegated to AI in just a few years if these technologies continue to evolve.
    • Document AI
      • https://cloud.google.com/solutions/document-ai
      • It extracts analytical information from documents and extracts structured data from unstructured data.
      • “Form Parser” announced
        • It extracts text and spatial structure from forms.
      • “Invoice Parser” annouced
        • It extracts various information from invoice as text and value.
      • [Opinion]
        Software such as OCR was the software that optical manufacturers were good at. However, we think that the technology will innovate with the full-scale entry of Google, which holds huge learning data of images and texts. And processing text data after extraction is Google’s specialty. This is because Google has a highly accurate natural language processing engine. Furthermore, with that as a hook, Google may approach all clerical work involving handwriting and provide ERP.
  • Solution for “Solve unique problems with AI”
    • E.g. Solutions for the retail industry
      • Recommendations AI 
    • E.g. Solutions for the financial industry
      • “Lending Document AI” announced
        • It is a new version of “Document AI” that can quickly process mortgages of financial institutions.
        • It can process documents about borrower income and assets.
      • “Procure-to-Pay Document AI” announced
        • It extracts structured data from specific documents such as invoices and receipts and helps automate the procurement cycle.
  • AI Platform
    • https://cloud.google.com/ai-platform
    • In order to generate value with AI, it is necessary for ML engineers, data scientists, and developers to work as a team. This platform supports the process of each person in charge.
    •  Services for ML engineers
      • AI Platform Pipeline
      • “Continuous monitoring service” announced
        • It is a service that can monitor the performance of machine learning models.
      • “ML metadata management service” announced
        • It is a service that can determine the origin of a model for training, debugging, auditing and collaboration.
    • Services for data scientists
      • Integrating Auto ML into AI Platform
        • It increases non-code and codebase options when building ML models.
      • AI Platform Notebooks generally available
        • It provides an integrated and secure JupyterLab environment for deploying ML models in experimental, development or production environments.
    • Services for developers
      • Pre-trained APIs
      • Updated image model of AutoML
  • Explainable AI
    • https://cloud.google.com/explainable-ai 
    • It is an explainable AI.
    • It is possible to see which features affect how much the model.
    • [Opinion]
      Deep Learning makes it possible to build highly accurate AI models by automatically determining the data and features to use. However, people tend to avoid what they do not understand. Therefore, there were cases where the AI model was not used even if it had high accuracy. If this service or technology can explain it, people may use it with peace of mind.

Sponsored Links